On Tue, Feb 3, 2009 at 7:47 AM, Roger Hui <[email protected]> wrote: > Why the restrictions against [ ] @ @: and constant fns? > (And yet [: is permitted.) >
[: would also be excluded; I'm sure John left it out only as an oversight. What John's looking for, I think, are "natural" examples where the dyadic fork applies. It is clear to me that there are many algebraic applications for the basic fork, as your tautological examples show. The Iverson approach involves expressing algorithms in an algebraic manner, so it looks like an interesting question to ask: What sorts of algorithms are nicely put as simple forks? The restrictions in question help us segregate algorithms that are a clean fit to the fork from those that have taken a more limited advantage of dyadic fork. My sense of it is that (] h f) is used far more often than (g h f). The dyadic fork seem mainly used to introduce additional data, once, into a sequence of transformations. The fact that the x parameter is made available twice, once at each of the outer tines, rarely seems to be of practical use. For comparison and contrast, I think Iverson's generalization of (v . u) can be readily shown to be applicable to an interesting range of problems. Is the dyadic fork at least as potent? If most uses involve [ or ] for one of the outer tines, perhaps we could find something where those need not be specified. (Both the monadic and dyadic uses of hook can be thought of as something along this line. I find hooks to fit naturally to computations very, very often.) -- Tracy ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
